convergence-tests.yml

name: "BioNeMo Model Convergence Tests"

on:
  workflow_dispatch:
    inputs:
      branch:
        description: "Branch to use"
        required: true
        default: "main"
        type: string
      model_config:
        description: "Model configuration to use"
        required: true
        default: "esm2_accelerate"
        type: choice
        options:
          - amplify_accelerate_te_fp8.yaml
          - esm2_accelerate
          - esm2_native_te_mfsdp
          - esm2_native_te_nvfsdp_thd
          - geneformer_native_te_nvfsdp_fp8

jobs:
  submit-lepton-jobs:
    runs-on: ubuntu-latest
    steps:
      - name: Checkout
        uses: actions/checkout@v4
        with:
          ref: ${{ inputs.branch }}

      - name: Set up Python
        uses: actions/setup-python@v5
        with:
          python-version: "3.11"
          cache: "pip"
          cache-dependency-path: |
            requirements.ci.txt
            ci/lepton/model_convergence/scripts/requirements.txt

      - name: Install deps
        run: |
          python -m pip install --upgrade pip
          pip install "hydra-core==1.3.2" "omegaconf==2.3.0" "leptonai" typer pandas
          python - <<'PY'
          import hydra, omegaconf, leptonai
          print("OK:", hydra.__version__, omegaconf.__version__, leptonai.__version__)
          PY

      - name: Submit Lepton Jobs
        env:
          LEP_LOGIN_CREDENTIALS: ${{ secrets.LEP_LOGIN_CREDENTIALS }}
        run: |
          lep login -c "$LEP_LOGIN_CREDENTIALS" || true
          pwd
          python ci/lepton/model_convergence/scripts/launch_job.py --config-name="recipes/${{ inputs.model_config }}" branch="${{ inputs.branch }}"